Direct Visualization of the Interfacial Degradation of Cathode Coatings in Solid State Batteries: A Combined Experimental and Computational Study

The interfacial instability between a thiophosphate solid electrolyte and oxide cathodes results in rapid capacity fade and has driven the need for cathode coatings. In this work, the stability, evolution, and performance of uncoated, Li2ZrO3‐coated, and Li3B11O18‐coated LiNi0.5Co0.2Mn0.3O2 cathodes are compared using first‐principles computations and electron microscopy characterization. Li3B11O18 is identified as a superior coating that exhibits excellent oxidation/chemical stability, leading to substantially improved performance over cells with Li2ZrO3‐coated or uncoated cathodes. The chemical and structural origin of the different performance is interpreted using different microscopy techniques which enable the direct observation of the phase decomposition of the Li2ZrO3 coating. It is observed that Li is already extracted from the Li2ZrO3 in the first charge, leading to the formation of ZrO2 nanocrystallites with loss of protection of the cathode. After 50 cycles separated (Co, Ni)‐sulfides and Mn‐sulfides can be observed within the Li2ZrO3‐coated material. This work illustrates the severity of the interfacial reactions between a thiophosphate electrolyte and oxide cathode and shows the importance of using coating materials that are absolutely stable at high voltage.


Introduction
Solid-state batteries (SSBs) have received considerable attention owing to their reduced flammability and the ability to use highenergy-density metal anodes compared with those of liquidelectrolyte battery systems. [1,2] Substantial progress has been achieved in the development of SSBs, with particular effort focused on identifying solid superionic conductors. Among LZrO coatings and thiophosphate SEs. [14,21] In addition, the relatively low oxidation limit of these ternary-metal-oxide coatings raises concern for their stability at high voltages. [14,21] An ideal coating material, therefore, requires a wide stability window that spans the cathode operating voltage and has low reactivity with both cathode and SE. Recently, with the aid of first-principles computation, lithium borates have been identified as a promising class of cathode coating materials that combine low reactivity with oxide cathodes and thiophosphate SEs with good oxidation stability. [14,22] Although computational screening has provided an efficient way to identify ideal compositions with targeted properties, experimental validation, and direct observations of how different coatings behave in a certain (electro-)chemical environments is crucial for the development of new coatings. Various methods have been established to investigate interfaces in liquid-electrolyte battery systems; however, obstacles remain in studying the interface behavior in SSBs, in part because of the difficulty of isolating an intact solid/solid interface in SSBs. [11] The characterization of the complex interparticle interfaces is further hindered by the nanosized length scale, air sensitivity, and amorphous nature of some decomposition products. [5,23] Advanced (scanning) transmission electron microscopy ((S)TEM) analysis with air-free sample preparation and transfer enables accurate probing of the chemical/structure evolution of the solid-state interfaces in SSBs. [7,11] In this paper, we combine different microscopy techniques (imaging and spectroscopy) to probe the chemical composition and microstructure of cathode composite interfaces with nanoscale resolution.
In this work, we select the lithium borate Li 3 B 11 O 18 (LBO) for detailed investigation because of its high oxidation stability limit and good chemical compatibility with both layered oxide cathode LiNi 0.5 Co 0.2 Mn 0.3 O 2 (NCM 523 ) and glassy Li 3 PS 4 (LPS) thiophosphate SE. Electrochemical methods were used to test the coating's effect on the discharge capacity and cycle-life of the SSB full cells. One commonly used ternary metal oxide coating, LZrO, which is predicted to be unstable at high voltage, was also characterized for comparison purposes to help identify the factors contributing to the coating stability. [14] Compared with the uncoated and LZrO-coated NCM 523 , the LBO-coated NCM 523 exhibits a greatly improved discharge capacity and capacity retention. The chemical and structural origin of the differences observed in the electrochemical behavior are interpreted using (S)TEM to observe the coatings and interfaces present in the cathode composite. Our combined computational and microscopic study helps to elucidate the critical coating parameters and provides a general methodology for future coating development and interface characterization in SSBs.

Density Functional Theory (DFT) Computation of Coating Stability
The electrochemical stability window of a material and the decomposition products at a given Li chemical potential can be predicted using a first-principles methodology previously developed. [5,24,25] The thermodynamic stability window represents the "safe" voltage range, over which there is no driving force for decomposition, whereas the oxidation limit at which a Li + ion and an electron can be topotactically extracted from a material represents the absolute upper voltage limit for kinetic stabilization. [25] Detailed descriptions of the computation methods are provided in previous work and in the Experimental Section. [5,24,25] The stability limits of LPS, LZrO, and LBO are summarized in Figure 1a and Table S1 in the Supporting Information.
As observed in Figure 1a, the stability windows of both LZrO and LBO overlap with that of LPS. However, the two coatings have divergent stabilities at ≈4.3 V versus Li/Li + . LZrO is predicted to become thermodynamically unstable at 3.42 V. Although the stability window can be kinetically expanded, LZrO will likely undergo decomposition within the common operating voltage window of NCM 523 cathode (e.g., 2.5-4.3 V). When charged to 4.3 V or higher, LZrO is likely oxidized by Li extraction via the predicted decomposition reaction expressed in Equation (1). In contrast, LBO exhibits a higher thermodynamic oxidation limit of 4.45 V due to the strong hybridization between boron and oxygen, [14] indicating that the LBO coating is likely to remain stable over the entire battery cycling voltage range of 2.5-4.3 V.
We also considered the chemical stability at the cathode/SE, cathode/coating, and coating/SE interfaces (Figure 1b). With no buffer layer, a high reactivity of −414 meV per atom is predicted between discharged NCM 523 and LPS, indicating a highly unstable interface upon direct contact. Both LZrO and LBO coatings exhibit significantly reduced reaction driving forces with NCM 523 (0 meV per atom for LZrO and −27 meV per atom for LBO). For the LPS/coating interface, a non-negligible driving force exists between the LZrO coating and LPS (−111 meV per atom) to form Li 3 PO 4 , whereas LBO is thermodynamically stable in contact with LPS. Therefore, LBO is expected to act as a chemically more robust barrier layer at the NCM 523 /LPS interface than LZrO, and provides a wide stability window spanning the voltage range of typical NCM 523 cathodes. A full list of the computed reactivities and the predicted reactions of all the reaction pairs is provided in Table S2 in the Supporting Information.

Effectiveness of LZrO and LBO Coatings
To experimentally investigate the stability and effectiveness of the two coatings, both LBO-coated and LZrO-coated NCM 523 were prepared using sol-gel coating methods as described in the Experimental Section. Electron microscopy images of the uncoated/coated NCM 523 particles are presented in Figure 2. As observed in the scanning electron microscopy (SEM) images in the top three panels of Figure 2, the morphology of the NCM 523 secondary particles is preserved after the coating process. TEM and high-resolution transmission electron microscopy (HRTEM) images indicate that the uncoated NCM 523 particle has a clean surface with lattice fringes extending to the particle edge (Figure 2b,c). As observed in the TEM and HRTEM images in Figure 2e,f (for the LZrO-coated NCM 523 ) and Figure 2h,i (for the LBO-coated NCM 523 ), a thin and uniform coating layer ranging from a few nanometers to ≈10 nm in thickness is observed on the surface of the NCM 523 particles. The HRTEM images in Figure 2f,i reveal amorphous features for both the LZrO and LBO coatings.
Full-cell SSBs using the uncoated/coated NCM 523 cathodes were constructed and cycled at a constant current (0.1 mA cm −2 for charge and 0.05 mA cm −2 for discharge) between 2.5 and 4.3 V versus Li/Li + to evaluate the effectiveness of the LBO and LZrO coatings. The cells were assembled in an Ar atmosphere using identical procedures; LPS powder from the same batch was used to maximize consistency. Figures S1-S3 in the Supporting Information show the representative charge-discharge curves of uncoated, LBO and LZrO-coated NCM 523 in the SSBs.  worst capacity retention (42.7% after 100 cycles). Both coatings improve the full-cell performance, providing larger capacity and higher retention. Compared with the LZrO coating, the LBO coating yields a significantly enhanced discharge capacity (146.5 mAh g −1 in the 1st cycle) and superior cycling stability (78.0% capacity retention after 100 cycles), suggesting a more effective protection of the NCM 523 /LPS interface. This observation is consistent with our DFT computation, in which LBO is predicted to exhibit excellent chemical compatibility and electrochemical stability at a high cutoff voltage of 4.3 V.
To better understand the electrochemical behavior of NCM 523 with different cathode coatings, (S)TEM was used to investigate the coating integrity and interphase formation during cycling. HRTEM images of the LBO-coated and LZrO-coated NCM 523 in composite cathodes before any electrochemical cycling are presented in Figure 4a,c, respectively. Both the LBO and LZrO coatings are amorphous and intact after mixing with LPS and carbon nanofibers (CNFs). The morphology and amorphous nature of LBO coating remain unchanged after the 1st cycle ( Figure 4b). In contrast, after the 1st cycle, the LZrO coating transformed from an amorphous layer into nanocrystalline phases with nonhomogeneous coverage on the surface of the NCM 523 particles, as observed in Figure 4d. As shown in the fast Fourier transform (FFT) image inserted in Figure 4d, the observed nanocrystalline phase can be indexed to the cubic ZrO 2 structure (space group Fm3̅ m, inorganic crystal structure database (ICSD) No. 53998). This phase transformation is likely the result of the Li extraction from LZrO under high voltage, consistent with the theoretical prediction expressed by Reaction (1). In fact, ZrO 2 nanocrystals can already be observed in the delithiated state even after initial charging to 4.3 V ( Figure S4, Supporting Information). The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image and energy-dispersive X-ray spectroscopy (EDX) mapping in Figure 4e After the LZrO coating breaks down, NCM 523 is likely to be subjected to direct contact with LPS, leading to the formation of a highly unstable interface. The HAADF-STEM image in Figure 5a shows the morphology of the LZrO-coated NCM 523 after the 1st cycle; the surface of NCM 523 is rough and mostly covered by island-like nanocrystals. As observed in the scanning transmission electron microscopy-electron energy loss spectroscopy (STEM-EELS) line scan spectra in Figure 5b,c, inhomogeneous distributions of Ni, Mn, and Co can be observed from the edge to the bulk region. Notably, although signals from Ni and Co are observed in the first few EELS line spectra in Figure 5c (corresponding to the edge of the NCM 523 particle), the signal from oxygen can barely be detected. This result suggests that Ni and Co no longer exist in the oxide form at the very surface and the elemental inhomogeneity is not due to the typical surface segregation and densification observed in layered lithium transition metal oxide cathodes in liquid-electrolyte battery systems. [26,27] Overlap of Ni, Co, and S is observed in the EDX mappings in Figure 5e, suggesting the possible presence of transition metal sulfides in the decomposition products.
In contrast to the phase decomposition and aggregation observed in the LZrO-coated sample, the LBO coating remains amorphous and intact even after extended cycling in an SSB. As observed in Figure 6a, the morphology and amorphous nature of the LBO coating are well preserved after ten cycles. A uniform distribution of the LBO coating on the NCM 523 particle after ten cycles is observed in the EDX mappings in Figure 6b,c. More images taken on the same sample are presented in Figure S5 in the Supporting Information. The HAADF-STEM image of the LBO-coated NCM 523 after ten cycles is presented in Figure 6d, with a clean surface of the coated NCM 523 particle observed in the interface region in contact with the LPS particle. With the protection of the intact LBO coating, no interfacial reaction between LPS and NCM 523 is detected in the LBO-coated samples. Homogeneous distributions of different elements are also found in the corresponding EELS line spectra of the LBO-coated NCM 523 sample in Figure 6d,e, indicating the chemical integrity of the NCM 523 particle. The atomic-resolution HAADF-STEM image in Figure 6f shows the crystalline structure of the LBO-coated NCM 523 after ten cycles, in which Adv. Energy Mater. 2020, 10,1903778  the bright and dark columns correspond to atomic columns of transition metal cations and Li-ions, respectively. Interestingly, the Z-contrast difference decreases in the edge region of the NCM 523 crystal compared with that in the bulk region ( Figure 6f and Figure S6, Supporting Information). The FFT image of the edge area reveals a rock-salt structure (inset of Figure 6f and Figure S6, Supporting Information), indicating that the surface structure of NCM 523 changed from a layered structure (R3̅ m) to a rock-salt structure (Fm3̅ m), similar to the surface densification reported in liquid-electrolyte battery systems. [27] While this may contribute to the observed capacity fade, the LBO coating remains stable with long-term operation as shown by the electron microscopy images of the wellpreserved LBO coating after 50 cycles ( Figures S7 and S8, Supporting Information).

Interfacial Reaction(s) between NCM 523 and LPS after the LZrO Coating Breakdown
To further investigate the NCM 523 /LPS interparticle interface and decomposition products after breakdown of the LZrO coating, HAADF-STEM and EDX analysis of the LZrO-coated sample after long-term cycling was performed. Figure 7 presents the STEM-EDX analysis of the NCM 523 /LPS particles after 50 cycles. The reaction between NCM 523 and LPS led to the formation of an extended interface region, as observed in Figure 7a. According to the elemental mapping in Figure 7c, the distributions of Ni and Co overlap well in the decomposition region but have become spatially separated from the distribution of Mn (Figure 7b, overlay map of Ni/Mn). The segregation of transition metals into distinct CoNiS and MnS domains is consistent with the computational reaction prediction (Reaction (2)), which indicates separate (Co,Ni)-sulfides and Mn-sulfides to form, though the STEM-EDX cannot identify the exact stoichiometry of the reaction products. Notably, some of the predicted decomposition products, such as Co(NiS 2 ) 2 , are likely to be electronically conductive. [22,[28][29][30] Thus, the electrochemical decomposition of LPS may not be passivated by the chemical reaction between NCM 523 and LPS, leading to a progressive degradation with long-term cycling and significant capacity fade. ( )

Discussion
While many cathode coating materials have been proposed and tested in SSBs, their effectiveness varies. While a lack of coating stability will influence battery capacity, it is difficult to identify specific mechanisms and severity of degradation from capacity loss with cycling. For this reason we performed a detailed microscopy study and correlated it with predictions that can be made from first principles calculations. The remarkable breakdown of the LZrO coating and the integrity of the LBO coating at high voltages provide strong experimental confirmation of previous theory results and are fully consistent with the idea that the high voltage limit of coating materials is determined by the extent of hybridization of the oxygen anion: [14] The strong covalent bonding in the BO 3 and BO 4 groups in Li 3 B 11 O 18 push the oxygen states down in energy, thereby protecting them from oxidation to a much larger extent than in LZrO. As revealed in this study, the rapid decomposition of the LZrO coating occurs due to Li extraction at high voltage, leading to both morphology and chemical changes of the coating. After Li is extracted, the thin LZrO coating transforms into aggregated regions of a poor Li conductor (ZrO 2 ), which further results in inhomogeneous coverage on the NCM 523 particles, exposing them to the LPS surroundings. Decomposition of the LZrO coating was found to continue upon cycling, consistent with the large capacity fade. While S/O exchange is predicted by DFT for an oxide/thiophosphate interface it is remarkable that even at room temperature segregation of transition metals Adv. Energy Mater. 2020, 10,1903778  into distinct CoNiS and MnS compounds can be identified as products of the reaction between NCM 523 and LPS. The agreement between ab initio computations based on a thermodynamic framework and the experimental observations indicates that even at room temperature near equilibration occurs on a very small scale. In fact, the segregation of Mn was also observed in a charged LiNi 1/3 Co 1/3 Mn 1/3 O 2 -LPS composite after being heated above 300 °C. [8] In that study, Tatsumisago and coworkers identified crystalline transition metal sulfides, such as MnS and CoNi 2 S 4 , as well as Li 3 PO 4 using synchrotron X-ray diffraction. [8] Our study shows that even at room temperature during normal battery operation the precursors of those crystalline degradation products already form. The oxidation of S 2− in thiophosphate electrolytes has also been observed in previous X-ray photoelectron spectroscopy studies. [31,32] It should be pointed out that while the key findings presented in this study are based on an SSB system, the electrochemical stability of a coating material is relevant in liquid electrolytes as well. LZrO will be prone to Li extraction under high voltage in both liquid and solid electrolyte. However, a significant difference is that in a solid-state battery the oxygen released from the LZrO decomposition can in turn exchange with the S 2− in the LPS surroundings, resulting in a decreased local oxygen partial pressure. Whether this low oxygen partial pressure is transmitted through the thin amorphous coating and leads to reductive instability of NCM 523 in a solid-state battery similar to densification in a liquid cell is unclear, but our result that even when the LBO coating remains intact the NCM 523 surface undergoes disordering similar to what is seen in liquid electrolytes, is suggestive. Understanding oxygen transport through cathode coatings should therefore be a topic of investigation.
Our results highlight the importance of electrochemical stability of coating materials in practical applications. An effective coating is important to protect NCM 523 particles from decomposition driven by its reactivity with the solid-state electrolyte. In addition to the (electro-)chemical stability and a reasonably good ionic conductivity, an effective coating should also exhibit good deformability to maintain its morphological integrity for a good contact and homogenous coverage throughout the entire cycle life. As shown in Figure S7 in the Supporting Information, the amorphous LBO coating remains intact on the NCM 523 particles after 50 cycles, allowing an effective protection of the interparticle interface. The soft nature of the lithium borate glass is expected to provide better mechanical "viscosity" on the NCM 523 particles than a mechanically stiff and brittle coating such as LiNbO 3 to accommodate the anisotropic volume expansion and contraction of NCM 523 during lithiation and delithiation. [13,33] The presence of a coating also brings new considerations into the cell assembly process and the full-cell fabrication. Ball-milling and/ or aggressive grinding during mixing of the coated NCM 523 and LPS tend to break the coated secondary NCM 523 particles and generate new unprotected interfaces. Gentle approaches (e.g., gentle grinding or soft milling) are suggested in the cathode composite mixing process ( Figure S9, Supporting Information).
Although direct observations from electron microscopy have provided valuable information on the composition, nanosized crystalline structure, and surface/interface morphology of the interfaces, isolating an intact solid interface and characterizing the spatial degradation of the amorphous electrolyte still remains challenging in SSBs. For example, although decomposition appears in a region of ≈200 nm in thickness in Figure 7a, the diffusion distance cannot be accurately determined as the spatial distribution of particles cannot be preserved during the Adv. Energy Mater. 2020, 10,1903778  sonication process of the sample preparation. Precise cutting of the buried cathode/interface cross-section is needed for better sample preparation. Combing varies microscopy technique, such as scanning transmission X-ray microscopy, together with analytical electron microscopy is also desired in future studies to probe how the electronic and chemical structure of the solid electrolyte interface evolve in dynamic electrochemical process. [34,35]

Conclusion
In this work, we combined ab initio computation with electron microscopy investigations to identify lithium borate LBO as a promising coating for the thiophosphate system and explain its superior performance. Both DFT calculation and microscopic characterization suggest excellent chemical and oxidation stability of the LBO coating in an NCM 523 -LPS cathode composite, effectively suppressing interfacial reactions, providing significantly higher capacity and enhanced capacity retention in a full-cell SSB. Direct observations from TEM and STEM reveal that LZrO is not stable at high voltages and is not effective for protecting the NCM 523 cathode. This work elucidates the critical parameters required for cathode coatings and provides insight into the design of coatings for high-voltage SSBs.

Experimental Section
Electrochemical Stability Calculation: The electrochemical stability window is calculated using the grand potential method developed in previous work. [5,24,36] All the DFT data are obtained from the Materials Project database. [37] The topotactic Li extraction voltage (kinetic stability limit) is calculated as the energy of removing one Li atom from a supercell with at least 16 formula units. [25] The DFT calculations were performed in Perdew-Burke-Ernzerhof generalized gradient approximation as implemented in the Vienna ab initio simulation package. [37,38,39] The interactions between ion cores and valence electrons are described using the projector augmented wave method with a plane-wave energy cutoff of 520 eV and a k-point density of at least 1000/n atom . [40] All supercells were fully relaxed in the DFT calculations.
The chemical reaction energy at the solid/solid interfaces (cathode/ SE, cathode/coating, and SE/coating interfaces) is calculated using the methodology developed by Richards et al. with DFT data obtained from the Materials Project. [5,37] The interface reaction calculation functionality is available through the Interface Reactions App of the Materials Project website. [37] Materials Synthesis and Coating Method: The coated NCM 523 samples were prepared using the sol-gel method reported in previous work. [41] For the LBO-coated NCM 523 sample (0.5 mol % LBO coated NCM 523 ), the LBO coating solution was first prepared by dissolving stoichiometric amount of Li acetate and triisopropyl borate in super dehydrated ethanol at 60 °C. Next, the NCM 523 powder was dispersed in the as-prepared coating solution with stirring. Then, the solvent from the flask was removed using a rotary evaporator in a hot water bath (60 °C) with ultrasound sonication, followed by a heat treatment at 350 °C. The LZrO-coated NCM 523 (0.5 mol%) sample was prepared using a similar process. Details of the material synthesis and coating process are reported in ref. [41].
Electron Microscopy Experiments: Sample preparation was performed in an Ar-filled glove box (water vapor <2 ppm and oxygen <0.1 ppm). The coated/uncoated cathodes were extracted from disassembled cells. The composite cathodes powders were diluted in hexane and sonicated to obtain good particle dispersion. The TEM samples were prepared by drop casting the solution onto a standard 400 copper mesh TEM grid with lacey carbon support. The samples were loaded into a Gatan 648 double-tilt vacuum-transfer holder to transfer the sample from the glovebox to the microscope in an inert Ar atmosphere. The HRTEM and STEM-EDX characterizations were performed using an FEI TitanX 60-300 microscope. For each sample, more than 20 interfacial regions have been examined from different LZrO or LBO coated primary particles. The electron beam currents were carefully controlled during imaging Adv. Energy Mater. 2020, 10,1903778  to prevent electron irradiation damage to the samples. The stability of NCM 523 /LPS interface before and after imaging were evaluated, with the HAADF-STEM images before and after EDX mapping shown in Figure S10 in the Supporting Information. The high-resolution STEM and EELS analyses were performed using the TEAM I microscope (a modified FEI Titan 80-300 microscope with double-aberrationcorrected (scanning) transmission electron microscope). SEM images were obtained on a Zeiss Gemini Ultra-55 analytical field-emission scanning electron microscope at the Molecular Foundry at Lawrence Berkeley National Laboratory (LBNL).
Electrochemical Characterization: Full cells using uncoated or LZrO/ LBO-coated NCM 523 as the cathode, LPS as the bulk SE, and graphite as the anode/reference electrode were constructed and cycled at a constant current (0.1 mA cm −2 for charge and 0.05 mA cm −2 for discharge). Both cathode and anode composites contained 60% active material, 35% LPS, and 5% CNF were prepared through gentle hand grinding. The cells were assembled under an Ar atmosphere using otherwise identical procedure and LPS powder from the same batch to maximize consistency. To prepare the full-cell SSBs, ≈100 mg of LPS powder was cold pressed into a loose "pellet" under small pressure. Then, 10 mg of the cathode composite powders and anode composite powders were carefully added to both sides of the LPS pellet before applying a pressure of ≈350 MPa for 5 min within an in-house-designed pressure cell (13 mm inner diameter) in an Ar-filled glovebox. A pressure of 5 MPa was applied to the cells during cycling. All the electrochemical tests were conducted at room temperature under Ar atmosphere.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.